Models and Data for Higher-level Strategies for Computer-Aided Retrosynthesis

Models and data for Higher-level Strategies for Computer-Aided Retrosynthesis- <code>datasets.zip</code>: The dataset curation pipeline in this project relies on classified and atom-mapped reaction data generated using the NameRXN software, which we are unable to release. We release the reaction and route datasets that were generated as a result of this pipeline. The resulting datasets are included in this file.<br>- <code>template_relevance_models_and_data.zip</code>: Contains the files necessary to deploy ASKCOS and run synthesis planning with all four single-step models used in this projects (i.e., .mar files for the four model, buyables file with price information). This zip file also contains the reaction splits, templates, and model checkpoints that are not necessary for deployment.<br>- <code>higher-level_consol_model_and_data.zip</code>: Contains the files necessary to deploy ASKCOS and run synthesis planning with just the higher-level single-step model (with template consolidation).<br><b>Update </b><b>(2025-11-18)</b><b>:</b> Added <code>uspto_original_with_patent_id.csv.gz</code>, which contains the same rows as <code>datasets/reactions/uspto_original.csv</code> (<code>datasets.zip</code>) with the patent id column (<code>patent_id</code>).